deep-research

Deep Research - Orchestration Command

You are executing the Deep Research workflow. You coordinate autonomous research through a map-reduce architecture: clarify intent, spawn parallel explorers, synthesize findings, and delegate report generation.

CRITICAL: Commands CAN spawn agents. You will spawn research-explorer agents for exploration and research-reporter for report generation. Do NOT delegate orchestration to another agent.

STATE-MACHINE

stateDiagram-v2
    [*] --> clarify
    clarify --> plan : intent_clear
    plan --> explore : plan_ready
    explore --> synthesize : exploration_complete
    synthesize --> report : synthesis_complete
    report --> [*] : done

On each phase transition, report via:

rp1 agent-tools emit \
  --workflow deep-research \
  --type status_change \
  --run-id {RUN_ID} \
  --name "Research: {brief summary of research topic}" \
  --step {CURRENT_STATE} \
  --data '{"status": "running"}'

• RUN_ID comes from the generated Workflow Bootstrap section

State Progression Protocol:

1. Report each --step with --data '{"status": "running"}' when you enter that state
2. For non-terminal states: move to the NEXT state when done (entering the next state implies the previous completed)
3. For terminal states (those with → [*] transitions): report with --data '{"status": "completed"}' and --close-run when the step's work finishes
4. On error, transition to the appropriate failure state in the graph

Example sequence:

--step clarify --data '{"status": "running"}'       # entering clarify phase
--step plan --data '{"status": "running"}'          # intent clear, entering plan phase
--step explore --data '{"status": "running"}'       # plan ready, entering explore phase
--step synthesize --data '{"status": "running"}'    # exploration done, entering synthesize phase
--step report --data '{"status": "running"}'        # synthesis done, entering report phase
--step report --data '{"status": "completed"}' --close-run      # report work finished, workflow done